2.1 Measurement setup and data processing

The experimental setup used for this study is shown in Fig. 1. An aerosol inlet with a PM₁₀ sampling head was installed on top of a measurement container at a height of 3.5 m a.g.l. Along a vertical tube (inner diameter of 2.5 cm) the aerosol was transported into the measurement container. Downstream horizontal tubes (inner diameter of 1 and 0.53 cm) distributed the aerosol to the instruments.

The total aerosol particle number concentration (N_CN) was measured by a condensation particle counter (CPC, TSI model 3010) which was operated at a total flow rate of 1 L min⁻¹. The CPC counts single aerosol particles between 10 nm (50 % of particles at this size are detected) and 3 µm. In parallel, the PNSD was measured by means of a scanning mobility particle sizer (SMPS, TSI model 3936 with CPC model 3025). The SMPS scanned aerosol particle mobility diameters from 13.6 to 736.5 nm with a time resolution of 5 min. Upstream of the SMPS an impactor with 1 µm cutoff diameter was installed and prior to the CPC and the SMPS the aerosol was dried using a diffusion dryer with silica gel. The airborne measurements of PNSDs on board the Polar 6 research aircraft were conducted by means of an Ultra-High Sensitivity Aerosol Spectrometer (UHSAS) that measured in a size range from 70 to 1 µm.

Figure 1Schematic depiction of the measuring setup implemented in the Tuktoyaktuk ground-based (container) station.

Download

The number concentration of cloud condensation nuclei (N_CCN) was measured using a cloud condensation nuclei counter (CCNc, Droplet Measurement Technologies (DMT), Boulder, USA). The CCNc is a continuous-flow thermal-gradient diffusion chamber which is described in detail in Roberts and Nenes (2005). The CCNc was operated as recommended by Gysel and Stratmann (2013) for polydisperse CCN measurements. By stepping the temperature gradient every 12 min the SS was varied between 0.1, 0.2, 0.3, 0.5 and 0.7 % at a constant total flow rate of 0.5 L min⁻¹. For consistency checks between N_CN and N_CCN 1 % SS was also applied. To ensure stable column temperatures, the first 5 min and the last 30 s at each SS setting were excluded from the data analysis. The remaining data points were averaged. A SS calibration of the CCNc had been done at the cloud laboratory of the Leibniz Institute for Tropospheric Research (TROPOS) prior to the campaign to determine the relationship between the temperature gradient along the column and the effective SS. Based on recommendations given in Gysel and Stratmann (2013) and Rose et al. (2008), ammonium sulfate particles were size selected using a differential mobility analyzer (DMA-type Hauke medium) and then fed into a CPC (TSI model 3010) and CCNc, which were operated in parallel. This was done for all SS values that were also applied during the measurement campaign. A size-dependent activated fraction was obtained by dividing N_CCN obtained at different sizes by the respective N_CN. By fitting the resulting curve using a sigmoid function, the critical diameter d_crit (where 50 % of all singly charged particles are activated) was determined. Applying the Köhler theory (Köhler, 1936), these d_crit were used to determine the effective SS based on the theoretical activation diameter of ammonium sulfate particles. The resulting effective SS values for this study are 0.11, 0.21, 0.31, 0.51 and 0.70 %, respectively. These calibrated values were used for further calculations, while the values reported from here on in the text are rounded values.

Figure 2Map of Alaska and the north-western part of Canada showing the position of Tuktoyaktuk and Inuvik together with two additional layers of the MODIS instrument (installed on TERRA, NASA). The sea ice extent layer shows the frozen ocean surface in pink. The corrected reflectance (bands 7, 2, 1) layer shows liquid water in dark blue or black. Ice on the surface or in the form of ice crystals in cold clouds will appear turquoise whereas small water droplets in warm clouds will appear white. A low-pressure system, which was relevant for the measurements in Tuktoyaktuk, and its corresponding fronts are marked. The map was created for 15 May 2014 using NASA Worldview (https://worldview.earthdata.nasa.gov/?map=-126.907471,36.373535,-117.415283,42.815918&products=baselayers,MODIS_Aqua_CorrectedReflectance_TrueColor~overlays,MODIS_Fires_All,sedac_bound&time=2012-08-23&switch=geographic).

Download

2.2 Inferring particle hygroscopicity

Whether an aerosol particle acts as a CCN depends on its size, chemical composition and the water vapour saturation in its vicinity. Köhler theory can be used to model the critical saturation ratio S_crit required for activation of a particle to a droplet (Köhler, 1936). The water activity, one term entering this theory, can be expressed based on a single parameter representation, using the hygroscopicity parameter κ (Petters and Kreidenweis, 2007). The κ values reported in this study were calculated as follows, assuming the surface tension of the examined solution droplets (σ_s∕a) is that of pure water:

with

d_crit is the critical diameter at which particles are just large enough to be activated to a droplet when exposed to a certain saturation ratio, S. M_w and ρ_w are the molar mass and density of water while R and T are the ideal gas constant and the absolute temperature, respectively. To derive d_crit, simultaneously measured N_CCN and PNSDs were used. It was assumed that the aerosol particles are internally mixed. From that it follows that at a given SS all particles become activated to droplets when their dry size is similar to or larger than d_crit. Hence d_crit is the diameter at which N_CCN is equal to the value of the cumulative PNSD, with the integration being done from the largest measured diameter of the PNSD to lower diameters. The thus derived d_crit can then be used, together with the corresponding S (i.e. the calibrated SS at which N_CCN was measured), to derive κ for the ambient particles, based on Eq. (1). The inferred κ values correspond to particles with sizes of roughly d_crit. The uncertainty in κ, which results from the uncertainties of the PNSD measurements and the SS of the CCNc, was determined by applying a Monte Carlo simulation (MCS) in a similar fashion as done by Kristensen et al. (2016). A detailed description of this method is given in Appendix A2.

For atmospheric particles, κ can range between almost 0 for insoluble (e.g. soot and some organics) and 1.4 for very hygroscopic (e.g. sodium chloride) particles (Petters and Kreidenweis, 2007). A generally good estimate for a continental κ of around 0.3 is given by Andreae and Rosenfeld (2008). Wex et al. (2010) report that the hygroscopicity of marine aerosol particles covers a broad range from several κ values below 0.1 up to fewer values of 1, with a dominating κ value of 0.45.

2.3 Measuring site and meteorology

All measurements presented in this study were performed during a period from 1 to 17 May 2014 at the outskirts of Tuktoyaktuk, a town of less than 1000 inhabitants in the Inuvik region of the Northwest Territories, Canada. Figure 2 shows a map of Alaska and the north-western part of Canada together with the sea ice extent layer and the corrected reflectance layer of MODIS (Moderate Resolution Imaging Spectroradiometer). Tuktoyaktuk is situated north of the Arctic circle (69^∘26^′ N, 133^∘01^′ W) on the shore of the Beaufort Sea and 5 m above sea level. The area around Tuktoyaktuk has a low population density. The nearest town with more than 1000 inhabitants is 130 km to the south (Inuvik). The Beaufort Sea is located directly to the north of Tuktoyaktuk, and it is typically covered with ice from October to June. The pink colour in Fig. 2 shows the extent of the sea ice on 15 May 2014. The area of the frozen sea surface covers the entire Beaufort Sea and extends to the south of the Bering Strait. Hence, it can be excluded that aerosol particles of marine origin are formed locally during the measuring period. The landscape surrounding the measurement station is comprised of flat Arctic tundra with a subarctic climate. The characteristic precipitation is less than 300 mm per year and the mean annual surface temperature is below 0 ^∘C . The time series of the meteorological parameters temperature, relative humidity, pressure, wind speed and wind direction (Fig. 3) give an overview with respect to the ambient weather conditions during the whole sampling period. The measured temperature shows an increasing trend typical for the transition from Arctic spring to summer. This transition is driven by the increase of the daily incoming solar radiation and leads to a change in sea ice and snow cover and, consequently, to a change of Arctic aerosol particle sources. During this transition from spring to summer, the polar front is moving towards the north, resulting in a more frequent passage of low-pressure systems as well as enhanced dynamics and mixing in the boundary layer of subarctic areas. This can be seen in the high variability of the measured temperature (from −10 up to 15 ^∘C) and the relative humidity (from 45 up to 95 %). Furthermore, sharp changes in all meteorological parameters indicate the passage of local low-pressure frontal systems (e.g. the pressure drop and the wind shift on 13 May that correspond to a cold front). Cold and warm fronts are marked with blue and red arrows in Fig. 3, respectively, indicating a fast air mass change. In Fig. 2 a low-pressure system, which was located over the Beaufort Sea, is visible due to the corrected reflectance layer. The corresponding warm front is also visible in the meteorological parameters of Fig. 3.

Figure 3Time series of temperature, pressure, relative humidity as well as wind speed and direction measured directly at the aerosol inlet at an altitude of 3.5 m above ground level covering the entire measurement period in May 2014. The passages of cold and warm fronts are marked with blue and red arrows, respectively.

Download

3.1 Overview of N_CN, N_CCN and PNSD data for the entire measurement period

Figure 4Time series of N_CN raw (grey) and N_CN (i.e. the filtered data, black) as well as N_{CN raw>150 nm} (light blue) and N_CCN,0.1 (dark blue). The coloured boxes mark the three periods of measurements that were used for further analysis. The arrows at the top indicate the four overflight times of the Polar 6 research aircraft.

Download

The measurements of all aerosol parameters recorded for this study were temporarily influenced by local anthropogenic sources. Local emissions from ground-based sources such as industrial combustion, oil heating and the generator for the container power as well as occasional air traffic led to intensive short-term peaks in the measured time series of N_CN, N_CCN at SS above 0.1 % and the PNSDs.

The grey line in Fig. 4 shows the raw time series of the measured total aerosol particle number concentration, where we will refer to these data using the parameter N_CN raw. Besides a clear baseline of concentrations between less than 100 and 1000 cm⁻³, peaks up to more than 10 000 cm⁻³ occur during the whole sampling period. These peaks had a typical temporal duration of 1 to 5 min. Consequently, the first step of the data analysis was the application of a filter routine to eliminate time periods where the measurements were affected by local pollution. The filtering procedure is described in Appendix A1. The black dots in Fig. 4 are the remaining N_CN data points. Note that especially during phases of high ambient pressure the local pollution is more intensive and the filter eliminates most of the data during these periods. Hence, on 2, 11–13 and 15 May almost no N_CN data points remain. These are time periods that directly follow the maxima in the pressure time series of Fig. 3. Typical for high-pressure systems are temperature inversions near the ground level that can trap local emissions and cause an enhanced influence of local pollution. Furthermore, Fig. 4 shows N_CCN measured at SS = 0.1 % (N_CCN,0.1) and number concentrations of particles larger than 150 nm (integrated PNSD, N_{CN raw>150 nm}) for the entire measurement period. For N_{CN raw>150 nm} and N_CCN,0.1 the filter procedure was not applied. Generally, N_{CN raw>150 nm} and N_CCN,0.1 show similar trends, and both do not show pronounced peaks as those seen for N_CN raw. This indicates that the observed aerosol particles that we related to pollution occurred in the size range below 150 nm. We will elaborate on this below. It is, however, also worth noting that N_{CN raw>150 nm} scatters much more than N_CCN,0.1. This larger scatter originates in the higher frequency with which N_{CN raw>150 nm} was measured. Over the whole measurement period a mean N_CN of 188 cm⁻³ (and a median of 199 cm⁻³) was observed when excluding the pollution periods. This is in good agreement with an Arctic long-term study by Tunved et al. (2013) who report monthly mean N_CN for May (observed at Mt. Zeppelin, Ny-Ålesund, Svalbard, from March 2000 to March 2010) being slightly above 200 cm⁻³. Generally they observed the highest concentrations between April and July, which can be traced back to aged anthropogenic Arctic haze aerosol earlier in this time period and to new particle formation later (Tunved et al., 2013). That these two kinds of aerosol also play a major role in context of the present study is discussed in the following two sections by using air mass back trajectories and the PNSDs. Figure 5 shows a comparison of the mean of all of those PNSDs which were detected during time periods that were marked as clean (PNSD_c, corresponding to times when N_CN is shown as black dots in Fig. 4) with the mean of all other PNSDs. The latter are those for which an influence of local pollution was assumed (PNSD_p). Also shown in Fig. 5 are error bars that indicate the range between the 25 and 75 % percentiles. Both PNSD_c and PNSD_p are bimodal with an Aitken mode below 100 nm and an accumulation mode above 100 nm. Above 150 nm, the accumulation modes of both are almost equal, whereas the Aitken mode of PNSD_p is more pronounced than that of PNSD_c. Similar influences of local emission on PNSD were found at an urban background station in Helsinki by Wegner et al. (2012). They observed modes between 10 and 40 nm in median urban PNSD caused by traffic, domestic and district heating, comparable to our result, albeit at higher concentrations. The observations by Wegner et al. (2012) support our assumption made earlier that the observed high peaks in N_CN raw originate from local pollution. It also demonstrates the usefulness of the applied filter (see Appendix A1). As for N_CN, PNSD_c also agrees well with the observations of Tunved et al. (2013). A PNSD shown in Tunved et al. (2013), representing the monthly median PNSD for May over a period of 10 years in an Arctic environment, shows the same characteristic as PNSD_c of this study as shown in Fig. 5. Both are bimodal with a distinct accumulation mode and a smaller Aitken mode. The variability of PNSD_c and the dependence on the air mass origin is discussed in Sect. 3.3.

After applying the filter on N_CCN and the PNSDs only three distinct periods (Period 1: 3 May 07:00 to 5 May 00:00; Period 2: 7 May 00:10 to 7 May 09:00; Period 3: 16 May 05:00 to 16 May 12:15) of evaluable data remained, as these parameters were measured with a lower temporal resolution and were thus more prone to be effected by local pollution than N_CN. The three periods are marked with a blue, green and red square in Fig. 4. This colour code is continuously used in the following figures. The arrows at the top of Fig. 4 indicate the four overflights of the Polar 6 research aircraft. For these times a comparison of ground-based and airborne PNSDs of different altitudes was done, which is discussed in Sect. 3.5.

Figure 5Comparison of the median PNSDs for times which were assumed to be clean (PNSD_c, purple) or polluted (PNSD_p, black). The thin vertical lines (same colour code) indicate the range between the 25 and 75 % percentiles.

Download

3.2 Identification of air mass origins and potential source regions

We applied two approaches to investigate the history of the measured air masses. First, the origin of the air masses of the three periods is identified by means of air mass back trajectories. Second, the potential source contribution function (PSCF), which is a residence time analysis that results in a probability field (Fleming et al., 2012), is applied to identify regions that potentially act as source regions for aerosol particles measured throughout the whole campaign.

3.2.1 Origin of sampled air masses of the three periods

To assess the origin of the air masses of the three periods we used the Lagrangian analysis tool (Sprenger and Wernli, 2015). The LAGRANTO backward trajectories were calculated based on analysis data from the European Centre of Medium-Range Weather Forecasts (ECMWF). The data used have a horizontal grid spacing of 0.5^∘ and 137 vertical hybrid sigma-pressure levels from the surface up to 0.01 hPa. Hourly 10-day back trajectories were started in the region around Tuktoyaktuk (69.43^∘ N, 133.00^∘ W) at a pressure level of 25 hPa below surface pressure. More specifically, we initialized trajectories at 13 receptor sites in the horizontal plane, accounting for the uncertainty introduced due to the relatively coarse horizontal grid and the release at an individual point. One receptor site was directly at the coordinates of the measurement station and 12 were around the station. The upper panel of Fig. 6 depicts the bundle of trajectories for the three time periods.

Two main air mass origins were observed. The air masses of Period 2 originated in the north-east of Canada (in the province Nunavut). During May air masses from this area are typically highly contaminated due to high winter and springtime aerosol particle burdens which can be observed all over the Arctic (Shaw, 1995). In the following these air masses are termed “accumulation-type” air masses. The air masses of Period 3 originated in the south-west of Tuktoyaktuk (Eastern Russia, Kamchatka and the unfrozen North Pacific). In the following, these air masses are named “Aitken-type” air masses, and the naming of the air masses will be explained in the next subsection, Sect. 3.3. Furthermore, the trajectories indicate that Period 1 includes both accumulation- and Aitken-type air masses. The distribution of these two air masses during Period 1 is visible in the lower panel of Fig. 6. The figure shows the number of trajectories per hour (for Period 1) that originated east (N_accumulation) or west (N_Aitken) of Tuktoyaktuk. It can be seen that during the first part of Period 1 (until 16:00 on 4 May 2014) the air masses in Tuktoyaktuk were a mixture of accumulation- and Aitken-type air masses. Until 3 May Tuktoyaktuk was influenced by an anticyclone with a maximum pressure of 1035 hPa. The low-pressure gradient of this anticyclone led to a low wind velocity and a baffling wind (see lower panel of Fig. 3), which caused an alternation between the two air mass origins at the beginning of Period 1. The second part of Period 1 is characterized by a decreasing surface pressure and a constant easterly wind. N_Aitken decreased as well, with a temporal shift of less than 1 day, which indicates that only accumulation-type air masses were present at Tuktoyaktuk.

Figure 6(a) Ten-day back trajectories of the three periods started in and around Tuktoyaktuk at a pressure level of 25 hPa above the surface pressure. (b) The total number of 13 trajectories per hour of Period 1 was split up into the number of trajectories that came from east (N_accumulation) or west (N_Aitken) to illustrate the alternation of the air mass origin.

Download

3.2.2 PSCF analysis

The PSCF is a receptor modelling method that is based on air mass back trajectories. Originally it was developed by Ashbaugh et al. (1985) and was previously applied in numerous studies, including high-latitude studies such as Dall'Osto et al. (2017b) (Antarctic) and Yli-Tuomi et al. (2003) (Arctic). This model is commonly used to identify regions that have the potential to contribute to high values of measured concentrations at a receptor site. In the present study, the NOAA HYSPLIT trajectory model was used to calculate hourly resolved 10-day back trajectories based on 1×1^∘ Global Data Assimilation System (GDAS) meteorological data. To account for uncertainties in back trajectory analysis, every hour a set of 15 back trajectories was calculated that was composed of five different plane locations (one exactly at the measurement station and four in close proximity around it) at three altitudes (100, 200 and 300 m above the surface level). We apply the PSCF model according to Hopke (2016) on hourly average N_CN values with the 75 % percentile as threshold value. The threshold value defines which N_CN value is considered as a high concentration. We calculated the PSCF on the basis of 5×5^∘ grid cells. To account for bad statistics in grid cells with a low trajectory density, a weighting function according to Waked et al. (2014) was applied. The spatial distribution of the PSCF of N_CN is mapped in Fig. 7. The map shows several areas of enhanced PSCF values that can be linked to potential source regions. One of these spots of enhanced PSCF values is located in central Canada, which potentially can be linked to biomass burning. To prove this possible connection, we used the MODIS Active Fire Product (MCD14ML) to display the active wild fire spots that occurred between 21 April and 17 May 2014 as magenta dots on the map. Due to the local proximity of these enhanced PSCF values and a spot of detected active fires south of them, it is possible to measure high N_CN values due to biomass burning in central Canada.

Also active fires detected in Siberia show coincidence with a spot of enhanced PSCF values. From a detailed analysis on a daily basis we see that this region of high PSCF values shows the occurrence of active fires throughout the whole time of analysis. Another spot in Siberia can be explained due to its proximity to Norilsk (red square in Fig. 7), which is considered to be an Arctic point source for emissions due to nickel mining and aluminum plants. Although improvement has been achieved since the 1980s, Norilsk still remains one of the largest sources of anthropogenic Arctic air pollution, mainly due to the emission of particulates and sulfur dioxide (AMAP, 2006). Another point source of anthropogenic Arctic emissions is the Prudhoe Bay Oil Field in North Alaska, marked with a red diamond in Fig. 7. Gunsch et al. (2017) and Kolesar et al. (2017) recently found the emissions of Prudhoe Bay Oil Field to cause increased N_CN values and have impact on growth events of nucleation and Aitken-mode aerosol particles. Our PSCF analysis results in a spot of increased PSCF values in the vicinity of Prudhoe Bay Oil Field. This indicates that Prudhoe Bay Oil Field emissions potentially lead to enhanced N_CN values measured in Tuktoyaktuk. The largest area of enhanced PSCF values is the area of the north-west Pacific. This region seems to overall cause relatively high N_CN values measured at Tuktoyaktuk, most likely due to marine emissions. A detailed discussion about the impact of marine emissions on the aerosol particles measured in Tuktoyaktuk is presented in the following section.

The above discussed PSCF results give a rough idea about the location of possible aerosol particle sources for our measurements at Tuktoyaktuk. However, in our case, the precision of the PSCF method is limited due to the small amount of data.

Figure 7Map showing the PSCF function of N_CN, calculated on the basis of concentrations exceeding the 75 % percentile. The colour bar indicates the PSCF value and the red dot, the diamond and the square indicate the position of Tuktoyaktuk, Prudhoe Bay (Alaska, USA) and Norilsk (Russia), respectively. The purple dots display the location of active fire spots that occurred between 21 April and 17 May 2014, detected by MODIS (Active Fire Product – MCD14ML).

Download

Figure 8(a) Thick lines show the median PNSD of Period 1 (blue), Period 2 (green) and Period 3 (red). The thin vertical lines (same colour code) indicate the range between the 25 and 75 % percentiles. (b) Comparison of the median PNSD of periods 2 and 3 with 10-year median PNSD of April and June observed by Tunved et al. (2013) on Svalbard. The PNSDs observed by Tunved et al. (2013) for April and June are comparable in shape with PNSD_c2 and PNSD_c3, respectively.

Download

3.3 PNSD of the three periods

Figure 8a shows the median of the PNSD_c of the three periods discussed in Sect. 3.2.1, together with the 25 and 75 % percentiles. They were computed to examine their dependence on the origin of the air mass. The PNSD_c of Period 1 (PNSD_c1) and Period 3 (PNSD_c3) are bimodal with an Aitken mode below 100 nm and an accumulation mode above 100 nm, whereas that of Period 2 (PNSD_c2) only shows the accumulation mode. The large variability we observed in the shape of the PNSD is typical for the transition period from Arctic spring to summer, i.e. higher variation of source regions during the Arctic summer. As described in more detail in the introduction, although there is a pronounced annual cycle in PNSDs in the Arctic, common features concerning PNSDs are shared across the Arctic (Freud et al., 2017), and we use PNSDs reported in Tunved et al. (2013) for comparison with our data in the following, as these were the first long-term data describing the annual cycle. PNSD_c2 is similar to the PNSD that Tunved et al. (2013) observed for March and April on Svalbard. A direct comparison of PNSD_c2 and the median April PNSD of Tunved et al. (2013) is shown in Fig. 8b. The monomodal accumulation-mode aerosol is characteristic for the Arctic haze, which mainly consists of POM and sulfate (Quinn et al., 2002). Single particle analysis of aerosol particles samples taken at the Zeppelin Observatory, Svalbard, that occurred before the transition to the Arctic summer showed a dominance of spherical organic-like particles in the submicrometer range with an Eurasian influence (Behrenfeld et al., 2008). These Arctic haze aerosol particles typically are well aged (Heintzenberg, 1980; Quinn et al., 2002). Due to the shape of PNSD_c2, and since the air mass of Period 2 has its origin in a region where conditions in May are still winterly, it is very likely that we observed a typical Arctic haze air mass. In contrast, PNSD_c3 is comparable to PNSDs that are reported by Tunved et al. (2013) for June and July. PNSD_c3 and the median June PNSD of Tunved et al. (2013) are depicted in Fig. 8b. In addition to the accumulation mode the bimodal summertime Arctic PNSD shows an Aitken mode which most likely originates from particles formed by new particle formation (Engvall et al., 2008; Wiedensohler et al., 2011). A common precursor gas for new particle formation is DMS emitted from oceanic phytoplankton. This precursor is known to be more abundant during the Arctic summer when the marine biological activity has its maximum. An indicator for the presence of DMS is its oxidation product, MSA (Quinn et al., 2007). MSA also could be directly detected as a component of the PM itself in remote marine background aerosol and in plankton bloom areas (Zorn et al., 2008). Quinn et al. (2007) report the concentration of MSA for several Arctic measurement stations (e.g. Barrow and Alert – Tuktoyaktuk is located between the two) during at least 7 years. The MSA concentration starts to increase in April and has two maxima during the summertime, where both maxima were observed in Alert and Barrow (Quinn et al., 2007). The later maximum occurs in July and August and is due to the local productivity of phytoplankton while the surface water is free of ice. The earlier maximum, which occurs around the time of our measurements, can be associated with long-range transport from marine source regions from the North Pacific (Li et al., 1993). This fits well with the source region we found for the air mass of PNSD_c3 and can explain the presence of the Aitken-mode particles. The minimum between the Aitken and the accumulation mode can be explained by previous cloud processing during which further material was added to activated droplets via aqueous phase oxidation. After the evaporation of cloud droplets this process creates the bimodal PNSD with the Hoppel minimum (Hoppel et al., 1994). In our case the Hoppel minimum can be found at around 90 nm. While cloud processing is a well-known process for gaining PM and growing particles to larger sizes, particles can also grow by generation of PM directly from the gas phase, as described recently for Arctic conditions in e.g. Willis et al. (2016), Burkart et al. (2017b) and Collins et al. (2017). The observed minimum in the PNSD occurs when new particle formation takes place either by adding small particles to an already aged air mass or by mixing of different air masses with one air mass containing aged and the other one newly formed particles, where one could come from aloft. It should also be mentioned that it was recently described in Gunsch et al. (2017) and Kolesar et al. (2017) that emissions from Prudhoe Bay oil field, which is located at the northern shore of Alaska roughly 700 km west of our measurement location, influenced Arctic PNSDs by adding both high concentrations of small particles and particulate mass to larger particles. In summary, there is a number of reasons that can add to the observed bimodality of the size distribution, but small, comparably newly formed particles will make up the observed Aitken mode in all cases. Sources of the precursor gases forming these particles will differ from spring to summer, as mentioned above (Li et al., 1993).

In Sect. 3.2.1 it is described that the back trajectories for Period 1 altered between the two source regions. This observation is supported by the shape of PNSD_c1, which suggests that both source regions contribute to the aerosol particles observed during this period. PNSD_c1 is bimodal (similar to PNSD_c3) but with a less pronounced Hoppel minimum and a distinct accumulation mode (similar to PNSD_c2). Due to the strongly alternating air mass origins during Period 1, the attempt to separate the two cases in PNSD₁ did not succeed, and the aerosol particle population reported for Period 1 in this study comprises a mixture between accumulation- and Aitken-type air masses. However, due to the absence of detailed information on the composition of the aerosol particles such considerations remain speculative.

Figure 9The time series of N_CN and N_CCN in panel (a), d_crit in panel (b) and the inferred κ values in panel (c). The error bars of N_CN and N_CCN show the standard deviation whereas the error bars of d_crit and κ show uncertainties that are determined by means of a Monte Carlo simulation. The blue data points show the unfiltered data corresponding to SS = 0.1 %. The other colours correspond to the filtered data (SS ≥ 0.1 %) of periods 1, 2 and 3, marked with the blue, green and red box, respectively.

Download

3.4 Critical diameter d_crit and hygroscopicity parameter κ

In Sect. 2.2 it is described how the critical diameter d_crit and the hygroscopicity parameter κ can be determined based on the measured N_CCN and PNSD. d_crit and κ were derived for (i) the whole measurement period using unfiltered N_CCN,0.1 and corresponding PNSD and (ii) the three selected periods described above, using filtered N_CCN at all SS and PNSD_c. Figure 9 shows the time series of N_CN and N_CCN (mean concentration with standard deviation as error bars) in the upper panel, d_crit in the middle panel and κ in the lower panel. Note that data concerning all SS are only available during the three selected periods. In Sect. 3.1 we describe that the pollution occurred in the size range below 150 nm. N_CCN measured at SS higher than 0.1 % are affected by local pollution due to the lower d_crit (d_crit is discussed in detail below) and thus are not analysed for the whole measurement period except the three periods. The uncertainties for d_crit and κ as given by the error bars were determined by the use of a MCS. A detailed description of this method is given in Appendix A2. (The uncertainties for d_crit are typically smaller than the symbol size.)

During the whole measurement period the unfiltered N_CCN,0.1 covers a range between less than 10 and 200 cm⁻³ with a median of 45.2 cm⁻³. The median d_crit values as well as the median concentrations for the filtered N_CCN are presented in Table 1. The corresponding inferred values for κ are representative for aerosol particles with sizes in the size range close to d_crit and therefore can be assigned to the modes in the PNSD. Therefore, κ derived for SS = 0.1 % displays the hygroscopicity of the accumulation mode as d_crit for SS = 0.1 % is in the neighbourhood of the maximum of this mode. κ values for SS = 0.2 and SS = 0.3 % are representative for the size range close to the Hoppel minimum, whereas κ values for SS = 0.5 and SS = 0.7 % are representative for the Aitken mode. The median κ values for SS = 0.1, 0.2, 0.3, 0.5 and 0.7 % are 0.18, 0.28, 0.21, 0.23 and 0.26, respectively. These values are summarized in Table 1. For periods 1, 2 and 3, the κ values averaged over all SS are 0.22, 0.23 and 0.26, respectively. We will discuss in the following how these κ values relate to the measurement uncertainty.

Figure 10a shows the probability density function of all κ values that are presented in the lower panel of Fig. 9. The blue line displays the median of the inferred κ values which is 0.23 and the red lines are the 5 and 95 % percentiles. The width between these percentiles, σ_s,all SS, amounts to 0.23. To check whether these inferred κ values allow a physical interpretation of the variability of κ, the MCS was used, as described in the Appendix A2. In short, using MCSs, the uncertainty for each κ value was determined separately based on uncertainties in d_crit, N_CCN and S_crit. This uncertainty was again expressed as the width between the 5 and 95 % percentiles. The separate widths were averaged, yielding σ_MC,all SS, which was determined to be 0.16. To resolve physically driven changes, σ_s,all SS should be significantly larger than σ_MC,all SS (at least twice as large). But ${\mathit{\sigma }}_{s,\mathrm{all}\mathrm{SS}}/{\mathit{\sigma }}_{\mathrm{MC},\mathrm{all}\mathrm{SS}}$ only amounts to 1.44, which indicates that 70 % ($=\mathrm{1}/\mathrm{1.44}$) of the variability in the observed κ values is related to measurement uncertainties of the PNSD and the SS in the CCNc. For corroboration, the same analysis was done based on a subset of all data. In Fig. 10b the κ values at SS = 0.1 % are displayed as a probability density function with a median of 0.19 and a width between the 5 and 95 % (σ_s,0.1) percentiles of 0.23. The width between the 5 and 95 % percentiles of the MCS (σ_MC,0.1, only for SS = 0.1 %) is 0.14 so that the ratio between both is 1.64. Hence 60 % of the variability in the observed κ values at SS = 0.1 % is related to measurement uncertainties. Summarizing, it can be stated that our observed κ values show no significant dependencies on the SS or the air mass origin that can be resolved with our setup and method.

Figure 10(a) Probability density function (pdf) of all inferred κ values (filtered values of all supersaturations) of the lower panel of Fig. 9. (b) Probability density function for κ values at SS = 0.1 % (inferred from the unfiltered CCN and PNSD data of Fig. 4). The blue line displays the median and the red lines show the 5 and 95 % percentiles of the probability density function.

Download

Kammermann et al. (2010) measured N_CCN and inferred d_crit and κ in a subpolar environment during a ground-based measurement campaign in northern Sweden for SS from 0.1 to 0.7 % in July 2007. They report κ values in the range of 0.07 to 0.21. Moore et al. (2011) and Lathem et al. (2013) report κ values from airborne measurements in Alaska (April 2008) and Northern Canada (June to July 2008), respectively. Both observed values between 0.05 and 0.3 within a SS range of 0.1 to 0.6 %. Lower κ are likely to indicate a higher organic fraction in these environments. Particularly for Kammermann et al. (2010), the lowest values can be explained by the local proximity to the Stordalen mire, which is known to emit biogenic precursors of organic aerosol particles. In a modelling study by Pringle et al. (2010), the annual mean κ values at the surface in the region around Tuktoyaktuk were approximately 0.3. Overall the κ values derived in this study are comparable to previously published values.

Table 1Median values of N_CCN, d_crit and κ for different SS. Values for SS = 0.1 % are calculated using the unfiltered data that cover the entire measurement period, whereas the values for SS = 0.2, 0.3, 0.5 and 0.7 % are calculated using the filtered data of periods 1, 2 and 3.

Download Print Version | Download XLSX

3.5 Comparison of height-resolved airborne and ground-based PNSDs

During the campaign four overflights of the Polar 6 research aircraft were performed. Overflights 1 and 2 were single overflights at a constant altitude of 600 and 200 m, respectively. During overflights 3 and 4 eight legs in altitudes between 300 and 3000 m were flown. The comparison of the airborne and ground-based measured PNSDs of the four overflights is shown in Fig. 11. Arrows in Fig. 4 indicate the times when the four overflights took place. For overflights 1 and 2 simultaneous measurements of filtered PNSD_c exist. For overflights 3 and 4 the closest filtered PNSD_c measurements have a temporal distance of 7 and 6 h to the time of overflight, respectively. Hence, for the comparison in case of overflights 3 and 4 the unfiltered PNSD_d measurements are used. The airborne PNSDs measured by means of an UHSAS were recorded with a time resolution of 1 s and extrapolated to standard pressure (1013.25 hPa). In Fig. 11 the UHSAS distributions are generally displayed as the median of 100 measured distributions. Additional bars that indicate the range between the 25 and 75 % percentiles are added to the distributions of overflights 1 and 2. PNSD_c and PNSD_d, which were measured at the ground, are shown for ambient pressure.

Figure 11PNSDs measured at the ground-based station and on the Polar 6 research aircraft during four overflights. Overflights 1 and 2 took place in 200 m and 600 m, whereas overflights 3 and 4 were profile flights at altitudes between 300 and 3000 m. Airborne measurements from 70 nm up to 1 µm were done using an UHSAS, while ground-based PNSD_cs and PNSD_ds of aerosol particle diameters from 13.6 up to 736.5 nm were measured using a SMPS as indicated in the setup of Fig. 1.

Download

For overflights 1 and 2, the ground-based PNSD_cs agree well with the airborne PNSD in the overlapping size range of 70 to 736.5 nm, where airborne measurements were carried out at 600 and 200 m, respectively. Vertical temperature profiles observed by radiosondes over Inuvik show temperature inversions at altitudes of 1500 and 700 m for overflights 1 and 2, respectively (not shown here). This indicates that during these two distinct time periods the ground-based measurements of PNSDs are representative for the atmospheric boundary layer.

For overflights 3 and 4, the measured PNSDs varied with respect to the particle number concentration and shape for the flights in different altitudes between 300 and 3000 m. The airborne PNSDs of overflight 3 show the same shape at all eight heights with a clear decrease of the number concentration at lower heights. The integration of the PNSD measured at 300 m (black line) gives a total particle number concentration of 24 particles per cm³, whereas it is 512 cm⁻³ at an altitude of 3000 m, i.e. 20 times higher. However the shape of the PNSDs does not change with height, as all distributions are monomodal with a maximum at approximately 150 nm. The ground-based PNSD_d in the size range above 150 nm agrees best with the airborne PNSD that was measured at 1200 m. At smaller sizes no comparison can be done, as the local pollution produces a large mode below 150 nm. The ambient temperature recorded at the Polar 6 aircraft during overflight 3 indicates a temperature inversion near an altitude of 2000 m. For further investigation back trajectories at altitudes of 1000, 2000 and 3000 m were calculated to investigate the history of air masses at different altitudes. The trajectories arriving at altitudes of 1000 and 2000 m show an air mass origin in the area of the North Pacific, comparable to Period 3 in Sect. 3.2.1. But the trajectory that arrived in 3000 m indicates an air mass origin in the central Arctic and over Greenland. Hence, the origin of the air masses and the relatively higher particle number concentration in the accumulation mode of the PNSD may indicate that the typical aged Arctic spring aerosol, which was observed during Period 2, is present above the temperature inversion. This aerosol could be mixed down to lower layers accompanied by a dilution process, but aerosol observed at the lower levels is likely mostly of a different origin. Overflight 4 shows that the airborne PNSDs also may differ in shape depending on the height. The PNSDs between 1750 and 3000 m are monomodal with a maximum between 100 and 200 nm. Also, a comparably high particle number concentration was measured at an altitude of 3000 m. The PNSDs at lower heights imply a second mode below 100 nm. This Aitken mode is also present in the ground-based PNSD_d, which fits the airborne PNSDs that were measured below 1200 m. The air masses above 1750 m show characteristics of the typical aged Arctic accumulation-type aerosol (comparable to Period 2) whereas the air masses below 1200 m seem to consist of aerosol of marine origin (comparable to Period 3). For overflight 4, two temperature inversions were recorded between 2500 and 3000 m. The temperature inversions and the different shapes of the PNSDs are indicative of the presence of different air masses during overflight 4, although air mass back trajectories that arrived at 1000, 2000 and 3000 m indicate an air mass origin over the North Pacific for all three heights.

Stone et al. (2014) explain that layering of Arctic aerosol, as we observed it during overflight 4, is a function of where the aerosol particle sources are located. Thereby the crucial factors are the different pathways of aerosol transport in the lower Arctic troposphere. The cold air of the lower Arctic troposphere is covered by surfaces of constant potential temperature and forms a dome over the Arctic (Law and Stohl, 2007). According to Stohl (2006) three transport pathways are possible: (1) low-level transport followed by ascent along the surfaces of constant potential temperature; (2) only low-level transport; (3) uplift outside the Arctic followed by transport in the upper troposphere and descent in the Arctic. It is likely that the aerosol particles we observed in the upper levels of overflights 3 and 4 were transported via pathway 1 or 3 whereas pathway 2 might be responsible for the occurrence of the bimodal PNSD below 1200 m during overflight 4.

Note that the altitude-resolved PNSDs presented here only represent a short snapshot in time. Hence, our observations do not describe how the transition from Arctic spring to summer affects the Arctic PNSD in the different lower layers of the troposphere. However, the measurements during all four overflights show that the ground-based PNSD is similar to the airborne PNSD of the lowest tropospheric layers. Therefore it can be rejected that local natural sources contribute significantly to our measurements during the observed time period, at least after removing signals from local pollution. It is more likely that aerosol particles or their precursor gases are advected via long-range transport from lower latitudes in different height layers and mixed down in the lower Arctic troposphere.